Speculation is something I can do quite well.
My understanding of similar fueling systems is that the pedal controls the driver requested torque. The requested torque is compared to the current torque. The delta torque is converted to a delta MAF, which adjusts the throttle angle target.
A weighted rolling average of the last few MAF sensor (or MAP) values gives you a steady-state estimate for air mass. If there has been significant change in the throttle angle (tip-in, tip-out) then some adjustment is made to account for transient changes in air mass. Changes in variable valve timing are also accounted for, along with vacuum sources (like brake boosters). Once they have estimated the air mass entering the cylinder, it's converted to volumetric efficiency (LOAD) by dividing by the displacement (or a similar maximum value).
Once they estimate LOAD, it's used for feedforward fuel mass calculations.
The fuel trim is applied here. After some filtering, the oxygen sensor feedback is converted from Lambda (derived from ppm of oxygen that represents "complete combustion") to Fuel/Air ratio (inverse of AFR). The difference between FAR and an adaptive target FAR (slowly adjusts for changing fuel and road conditions).
Alot of stuff plays into this part. Status of the charcoal canister, estimated oxygen stored in the catalyst (this is actually a really hot topic right now), and expected perturbations will all be factored in. Each manufacturer is different here. Delta FAR is multiplied by the LOAD or air mass flow rate. Both yield similar scaling at the end of it all. The % change in fuel mass is calculated here (that's fuel trim).
The feedforward fueling is adjusted by the fuel trim and the total mass (calculated on a per cycle, per radian, or per cylinder basis) is then split up for each injector installed. If the DI system uses split injection, this will get calculated here.
A faster control loop is typically used here since injector pulses need to be pretty stinkin' accurate. The target fuel mass for each injector is converted to a pulsewidth based on fuel rail pressure, battery voltage, and other stuff. This ideally is a linear function, but typically high fuel mass will locally deplete the fuel rail so they increase the open time under heavy load.
As @
arghx7 said they may set an "end of injection" time or a "start of injection" time and the injection timing and length are scheduled for the next revolution. The next time it's that injector's turn, it will trigger a short on the injector solenoid as specified. If all goes well, the perfect amount of fuel will be injected, but most likely there will be feedback from the oxygen sensors and changing driver torque demands.
Sorry for not answering your question at all... I enjoy learning about this stuff.
My guess is that the fuel trim is applied the total fuel mass. The adjusted amount gets split up as mentioned above.